Merge pull request #312 from VikParuchuri/dev

Bugfixes

marker/layout/layout.py

@@ -73,7 +73,7 @@ def annotate_block_types(pages: List[Page]):
 
         for i, block in enumerate(page.blocks):
             if block.block_type is None:
-                block.block_type = "Text"
+                block.block_type = settings.DEFAULT_BLOCK_TYPE
 
         def get_layout_label(block_labels: List[str]):
             counter = Counter(block_labels)

marker/postprocessors/markdown.py

@@ -160,7 +160,7 @@ def merge_lines(blocks: List[List[MergedBlock]], max_block_gap=15):
                 text_blocks.append(
                     FullyMergedBlock(
                         text=block_surround(block_text, prev_type, prev_heading_level),
-                        block_type=prev_type,
+                        block_type=prev_type if prev_type else settings.DEFAULT_BLOCK_TYPE,
                         page_end=False
                     )
                 )
@@ -186,7 +186,7 @@ def merge_lines(blocks: List[List[MergedBlock]], max_block_gap=15):
             text_blocks.append(
                 FullyMergedBlock(
                     text=block_surround(block_text, prev_type, prev_heading_level),
-                    block_type=prev_type,
+                    block_type=prev_type if prev_type else settings.DEFAULT_BLOCK_TYPE,
                     page_end=True
                 )
             )
@@ -197,7 +197,7 @@ def merge_lines(blocks: List[List[MergedBlock]], max_block_gap=15):
     text_blocks.append(
         FullyMergedBlock(
             text=block_surround(block_text, prev_type, prev_heading_level),
-            block_type=block_type,
+            block_type=block_type if block_type else settings.DEFAULT_BLOCK_TYPE,
             page_end=False
         )
     )

marker/settings.py

@@ -71,6 +71,7 @@ def TORCH_DEVICE_MODEL(self) -> str:
     BBOX_INTERSECTION_THRESH: float = 0.7 # How much the layout and pdf bboxes need to overlap to be the same
     TABLE_INTERSECTION_THRESH: float = 0.7
     LAYOUT_BATCH_SIZE: Optional[int] = None # Defaults to 12 for cuda, 6 otherwise
+    DEFAULT_BLOCK_TYPE: str = "Text"
 
     # Ordering model
     SURYA_ORDER_DPI: int = 96

marker/tables/table.py

@@ -36,19 +36,19 @@ def get_table_boxes(pages: List[Page], doc: PdfDocument, fname):
         pnum = page.pnum
         # The bbox for the entire table
         bbox = [b.bbox for b in page.layout.bboxes if b.label == "Table"]
-
-        if len(bbox) == 0:
-            table_counts.append(0)
-            img_sizes.append(None)
-            continue
-
         highres_img = render_image(doc[pnum], dpi=settings.SURYA_TABLE_DPI)
 
         page_table_imgs = []
         page_bboxes = []
 
         # Merge tables that are next to each other
         bbox = merge_tables(bbox)
+        bbox = list(filter(lambda b: b[3] - b[1] > 10 and b[2] - b[0] > 10, bbox))
+
+        if len(bbox) == 0:
+            table_counts.append(0)
+            img_sizes.append(None)
+            continue
 
         # Number of tables per page
         table_counts.append(len(bbox))

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "marker-pdf"
-version = "0.3.3"
+version = "0.3.4"
 description = "Convert PDF to markdown with high speed and accuracy."
 authors = ["Vik Paruchuri <[email protected]>"]
 readme = "README.md"